Claude-skill-registry fusion-algorithm

Compare multiple algorithm options and pick the best; use when choosing or optimizing an algorithm.

install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/fusion-algorithm" ~/.claude/skills/majiayu000-claude-skill-registry-fusion-algorithm && rm -rf "$T"
manifest: skills/data/fusion-algorithm/SKILL.md
source content

Fusion Algorithm

Overview

Use an F-thread: three workers in separate git worktrees plus a judge queen. Each worker implements the algorithm differently. The queen compares and selects a winner.

Inputs

  • Algorithm description

Workflow

  1. Verify
    git
    and
    mprocs
    .
  2. Create session variables:
    • SESSION_ID
      ,
      BASE_BRANCH
      ,
      PROJECT_ROOT
      ,
      WORKTREE_ROOT
  3. Create
    .hive/sessions/<session-id>
    and
    WORKTREE_ROOT
    .
  4. Create worktrees:
git worktree add "{WORKTREE_ROOT}/impl-a" -b fusion/{SESSION_ID}/impl-a
git worktree add "{WORKTREE_ROOT}/impl-b" -b fusion/{SESSION_ID}/impl-b
git worktree add "{WORKTREE_ROOT}/impl-c" -b fusion/{SESSION_ID}/impl-c
  1. Write
    tasks.json
    and worker prompts.
  2. Write
    .hive/mprocs.yaml
    and launch mprocs.

tasks.json Template

{
  "session": "{SESSION_ID}",
  "created": "{ISO_TIMESTAMP}",
  "status": "active",
  "thread_type": "F-Thread (Fusion)",
  "task_type": "fusion-algorithm",
  "algorithm": {"description": "{ALGO_DESC}"},
  "base_branch": "{BASE_BRANCH}",
  "worktrees": {
    "impl-a": {"path": "{WORKTREE_ROOT}/impl-a", "branch": "fusion/{SESSION_ID}/impl-a", "worker": "worker-a"},
    "impl-b": {"path": "{WORKTREE_ROOT}/impl-b", "branch": "fusion/{SESSION_ID}/impl-b", "worker": "worker-b"},
    "impl-c": {"path": "{WORKTREE_ROOT}/impl-c", "branch": "fusion/{SESSION_ID}/impl-c", "worker": "worker-c"}
  }
}

Worker Prompt Outline

  • Worker A: clean, readable implementation
  • Worker B: alternative approach or data structure
  • Worker C: performance-focused implementation

Queen Prompt Outline

  • Compare correctness, complexity, readability, and test results
  • Pick a winner or merge best elements

mprocs Launch

mprocs --config .hive/mprocs.yaml

Output

  • Three implementations in worktrees and a final evaluation